AI is leaving the Box - it hears, it sees, it acts.

00:00:00: This is the first time anywhere in the world.

00:00:02: And this happened twice in China, and last few weeks where courts actually up you know essentially companies were sued because The reason for firing somebody was that they save money by using AI?

00:00:14: And the court said That's illegal for you to do that and it wasn't forced.

00:00:18: and now the guy has to get hired back.

00:00:23: Hello!

00:00:24: Welcome to our big bank tech report.

00:00:27: I'm very happy To say hello to Alvin who is on the show.

00:00:32: and where are you right now, Alvin?

00:00:34: Where in the world.

00:00:35: Back back in Seattle.

00:00:36: so looking forward to our conversation.

00:00:38: but I'll be on the road for the next three four weeks again soon.

00:00:41: So uh if you won't be seeing this background after a while.

00:00:46: So

00:00:46: you are traveling a lot and yeah, you're spreading your word.

00:00:50: And right now I have the impression that we are waiting for summer... ...and summer is coming very quick but we do not have the clothes.. ..we don't have shorts or scant screen lotion all this stuff….

00:01:04: …and i have

00:01:05: same impressions

00:01:06: with AI.

00:01:08: Right Now We talk alot about it About Agents That They Will Come!

00:01:12: And in Germany they said ok We have a wake-up call and next level AI is coming.

00:01:18: It can see, it can hear... ...it can plant checks and so on.

00:01:24: And where are we right now in your opinion?

00:01:30: I think that the technology's just progressing too quickly.

00:01:34: Every week there's new breakthrough.

00:01:37: You know, it used to be Open Claw now that the Hermes agent is leading open router token leaders.

00:01:47: So these things are just moving too quickly and its hard for industry in world to adapt.

00:01:55: I've got an example yesterday for them that AI is leaving the box.

00:02:01: That AI has not only a chatbox anymore, how we are using it if you need to speak with or talk to a chat GBT and giving us this speech.

00:02:10: And yes today i talked with a guy from Siemens.

00:02:14: Are you familiar with Siemens?

00:02:15: You have heard of Siemens.

00:02:16: Yes It's right now The biggest company in Germany as stock exchange after SAP and before it was Bayer, but then they bought Monsanto in America.

00:02:28: Now they have had a crash or something like that.

00:02:30: And Siemens guy told me that In the assembly line of great car factories there is right now AI Is looking at their assembly line?

00:02:42: It's working A lot and this shifting work in another way much more productive And that's maybe a reason, too.

00:02:51: That the stock exchange is quite doing well and even if we have crisis everywhere on the world.

00:02:58: all prices are high The people can't buy anymore so much but something is changing.

00:03:04: You have to feeling there was big change right now.

00:03:08: Yeah I think this technology has been integrated into all sectors And so whether you're talking about manufacturing or services, it's all coming.

00:03:19: It will make the world more efficient and productivity go up.

00:03:23: Now what is the labor impact of this?

00:03:29: The long term economic impact also?

00:03:32: how does that affect?

00:03:33: margins and profitability in these companies may help a few people But then that may also be temporary.

00:03:44: That's more or less, I think the mega trend we see and above the mega-trend a lot of news.

00:03:51: for example Elon Musk versus Open AI.

00:03:54: what is going on there?

00:03:59: The last couple weeks was big news.

00:04:01: A lot people thinks it as more drama between two people but really it's a redefinition of non-profits and its relationship to the economy.

00:04:14: Because what they're really suing for is can a company who started out as a nonprofit become later, or for profit taking all their money that people donated for public good into something that goes in somebody pocket?

00:04:27: So I'm not fan either of these people right now because i feel like you know there are doing maybe for ulterior motives but decision this case is going to have long-term impact.

00:04:40: So we need to watch it carefully.

00:04:42: But right now in Germany, nobody's talking about ChetGBT.

00:04:45: You know what?

00:04:46: All the people are talking about Anthropic.

00:04:49: Why do they use that?

00:04:50: It's incredible What this company is doing Right now, right?

00:04:54: Yeah, Anthropics has been doing some amazing things In terms of particularly started out really last year with Cloud Code And when product came out.

00:05:04: It essentially allowed people to have real productivity gains in coding and later on other areas, they've been really shipping products every week And the capabilities of their models were for a time best-in-the world.

00:05:20: I think over the last few weeks ChatGP has actually caught up.

00:05:24: The five point version of GPT had really... In some ways surpassed the capabilities of the latest anthropic models, because Anthropic actually held back their biggest model for either security reasons or some people say it's because they just don't have the compute capacity.

00:05:47: So I think this is also what we talked about before that there's no long-term winner and things are going to go back and forth.

00:05:55: And this race really right now US and China.

00:05:59: The real race right now that people are pushing for is between the U.S labs themselves.

00:06:05: But about their compute capacity, they have bought where?

00:06:08: Right now do you know it?

00:06:10: Oh yeah so Elon Musk just gave access to two hundred thousand ships to Anthropik And its actually going be a big help because I've been using cloud recently.

00:06:22: You run into your limits very quickly.

00:06:25: So now they've actually doubled their limits and I think most people are not Relatively satisfied with the capabilities both to speed of progress as well.

00:06:34: As you know Just how much capacity for tokens that you have on this subscriptions.

00:06:40: so but I Actually want to kind of dive into this a little bit because mostly mostly things like that's a commercial Agreement, but at the end of the day actually it helps.

00:06:50: It helps Elon more than actually it helped anthropic.

00:06:54: Yeah, because I don't know if you know but the XAI team and offering is really getting very little market traction.

00:07:02: So all of those chips were actually unused.

00:07:06: in fact recent reports came out that they were only eleven percent utilized.

00:07:10: so essentially ninety per cent a time.

00:07:12: it's just sitting there And for Elon we have these billions of dollars.

00:07:17: equipment are not being used.

00:07:19: It's actually waste now person paying them for these chip capabilities.

00:07:28: And given that he's about to go IPO soon with SpaceX, which now owns XAI this is a revenue stream that he needs and then also more importantly it because has big grudge against open AI as you know.

00:07:42: so helping your enemy is just good at hurting your enemies.

00:07:48: So his doing in the way of punishing Open AI.

00:07:53: But lastly, it actually also is a way to control Anthropic because now Anthropics dependent on Elon Musk's mood On things Because he He Now controls this big you know one of the big portion of supply for for compute.

00:08:11: so You Know all know It's A bigger win for E-Line that most people realize.

00:08:16: And then there is another player in the field, it's a Pentagon!

00:08:19: What's going on with the Pentagon right now?

00:08:22: Why are they so interested in

00:08:24: AI?".

00:08:25: Yeah

00:08:26: I mean i think ever since the mythos released The Pentagon has seen AI now elevate and its awareness to say hey this something we have an advantage of.

00:08:37: You know, they got into a bit of a fight with anthropic over access and control about that solution.

00:08:46: And so now they're going back to all the other companies.

00:08:48: uh...and they just signed seven contracts with everybody except for in Prophec For kind-of next gen AI solutions So integrating them into military services.

00:09:01: All those companies have agreed to essentially not influence The Department Of War Over how that technology is being used.

00:09:13: And then, you know what?

00:09:14: My favorite company Google has had some news in the health field!

00:09:21: What's going on there?

00:09:22: Is it an official news or was it somebody who gave it to The Press?

00:09:27: I think they released their cold clinician product and also just released their Omni model Gemini Omni.

00:09:34: So this actually been a couple of weeks since its pretty busy for them And, you know this is really about integrating AI into industry and health care's one of the biggest industries in the world.

00:09:46: So can you

00:09:46: explain what's going on there?

00:09:47: What is...what's going On In The Health Field?

00:09:50: to understand that everybody Can Understand That I go To The Doctor And Then

00:09:54: Yeah so essentially This Becomes The Guide To The Doctors Or They Can Make Better Decisions Right.

00:10:02: Years Ago There Was Something Called Open Evidence That Came Out In The US and something like sixty to seventy percent of doctors right now are using it, essentially is a custom AI engine just for medical cases.

00:10:13: It's only trained based on real medical studies And so you avoid a lot of hallucination issues with general-based AI systems.

00:10:25: I think that what they're doing is following the model saying we'll make a custom vertical version specifically for doctors.

00:10:36: the hallucination and everything that's in there can be cited, attributed to some real study.

00:10:42: And I think this gives confidence for clinicians or doctors... ...to apply these answers in a way which will affect lives.

00:10:52: So is it a system for the doctor?

00:10:54: For the patient?

00:10:55: of both?

00:10:56: It's mostly for doctors, actually.

00:10:58: For doctors to use and advise them.

00:11:01: so it is an augmentation system not a replacement system.

00:11:05: Now open AI did come out maybe few months ago with their kind of more consumer oriented systems To say hey you don't need the doctor or talk to your AI And I think they are using this as way to create more traffic.

00:11:22: So these are different approaches.

00:11:24: Honestly, I think right now the Google approach is probably more responsible because you know have a doctor that will essentially filter DAI and then help disseminate what's useful whereas when everybody thinks there're doctors sometimes they may make some poor health care decisions.

00:11:42: There was this big discussion every time in the health field in the radiology field, you need more doctors than ever because there are more scans then with machines and they have to explain what's going on.

00:12:04: He said it is opposite that AI needs more jobs but not less.

00:12:09: so your opinion?

00:12:12: I think sounds nice.

00:12:14: saying things like this Kind of job displacement that this makes people feel better.

00:12:22: Yeah, and I think there are certain areas where this might happen.

00:12:26: but it's also because they were so few Radiologists That that that's the case.

00:12:32: But the issue is that?

00:12:34: You know these AI systems are not just attacking up one particular segment.

00:12:38: It goes across the board on all white collar labor.

00:12:42: And then there will not be more replacements for all of these areas, right?

00:12:48: Because not everything is so understaffed as the medical area.

00:12:53: It takes a decade to get your medical degree often and going through all residency in every thing.

00:13:00: So now everybody wants you put it at that time.

00:13:03: But now every lawyer, every consultant, every accountant Every office worker isn't exposed to AI.

00:13:10: How do say there's gonna be more than those things?

00:13:14: So talking about efficiency, we have to shift our discussion to China because that's land and of efficiency.

00:13:23: And there are some news about DeepSeek on WhatsApp

00:13:28: Q.

00:13:29: I've never heard it before but you mentioned it before.

00:13:32: please explain what is going in China.

00:13:35: Yeah, so I think you know recently deep-seekers is now seeking to have their first round of funding.

00:13:41: They were always privately funded before by the founders.

00:13:44: So they never raised around them.

00:13:45: But this was our first round a funding and they're going to raise something Around seven billion dollars which is the largest private raise ever in China And we'll value them around fifty billion dollars.

00:13:56: The funny thing is when they started to raise about two months ago They were valued around ten billion and so they've gone up from ten to fifty billion because now everybody wants a piece of them And it's going to be invested by large corporates like Alibaba in Tencent, but also Some the state own funds that used to be targeted towards semiconductors.

00:14:18: So I think this allows.

00:14:19: essentially the government is saying hey.

00:14:20: I want a piece Of the action and i want To You know make this have some level control as well right?

00:14:31: kind of similar to what's happening actually in the

00:14:32: U.S.,

00:14:33: if you look at OpenAI and Anthropoc, they're also invested by Google and Microsoft here.

00:14:39: so it is a mirroring but when does shows that companies were in the past essentially non-revenue focus will probably become revenue focused because their investors are going to care.

00:14:54: And subcube?

00:14:55: What about subcubed?

00:14:57: Yeah, subcube is actually an American company or it's done by an American lab and essentially its a way to solve some of the issues associated with transformers.

00:15:07: So transformers are what all other major large language models based on but they attention mechanism essentially requires very large memory footprints to be able operate.

00:15:22: And so this allows you to, they're essentially saying don't focus on everything.

00:15:27: Don't pay attention to everything only Pay attention to the most important things and They can reduce essentially memory usage by like fifty X. So if that's the case then You don't need these giant Server racks To run maybe running at home On laptops or even phones in future.

00:15:46: That's not a good news for NVIDIA.

00:15:50: I mean it's good and bad, right?

00:15:53: We've been talking about this for a while.

00:15:55: In the long run more AI inference will move to edge which is great because then everybody gets access in the large server racks, because you need just a big amount of data and a lot of systems to build up the original core frontier models.

00:16:25: But then they can be distilled into some More optimized models that will run everywhere.

00:16:30: So it's a kind of good and bad at the same time.

00:16:34: And now we have something like a showstopper, I think in every show We are talking about AI is replacing jobs so white-colored jobs and so on.

00:16:45: then there isn't news from China That There Is A Court and said okay or we can do It?

00:16:57: Yeah, I mean you know.

00:16:58: I think this is a concern that's happening everywhere in the world But this is the first time anywhere in the World and this happened twice in China in the last few weeks where Courts actually up You know essentially companies were sued because The reason for firing somebody was that they saved money by using AI And the court said That's illegal For you to do that?

00:17:18: And it Was enforced and now the guy Actually has To get hired back.

00:17:28: So I think this actually is a good precedent, because it shows that human dignity can be preserved even in an age where efficiency is the primary objective of companies.

00:17:43: And I hope other countries maybe take some learnings from this.

00:17:49: if we don't somehow protect the dignity There will be a major uprise.

00:17:56: Essentially people in the streets because there's going to mass layoffs and those people are going to be frustrated, have no other means to survive.

00:18:06: so we need find ways of creating smooth transition.

00:18:09: So were you astonished that this news came from China not England or America even Germany?

00:18:18: Not really.

00:18:19: I don't know if many don't realize, but China actually is the most heavily regulated AI market in the world right now.

00:18:29: They enforce more than a dozen policies that keep... That slows down their own AI industry and way to create either higher public good or social stability or copyright rules or reducing AI safety risks.

00:18:50: Right, so they're doing the things that I think the world should be doing but actually putting it out and enforcing.

00:18:55: And Europe has the EU Act But doesn't really come into effect until August this year.

00:19:01: So China's been doing for last two or three years Keeping making sure its not an addictive system Making everything transparent when you use AI and is putting materials that have to properly labeled.

00:19:15: Liability of a company can be directly affected if bad things happen due to their models.

00:19:20: That's not happening anywhere else in the world, and so for them too say hey uh the effects on labor um is important.

00:19:29: that comes from technology come from AI.

00:19:31: I think it's really consistent with what they've been doing.

00:19:36: We started in the

00:19:38: U.S.,

00:19:38: and we switched over to China, so let's talk a little bit about Europe or especially about Germany.

00:19:45: tomorrow I have an interview with A guy from Bavaria.

00:19:49: you've ever heard of Bavarian?

00:19:51: Of course

00:19:52: uFS beer And all this stuff You know but the Bavari is a silicon state in Germany is a Silicon Valley of Germany and the minister tomorrow is here.

00:20:01: and How do you see Germany in this race.

00:20:06: Are we outside of the race?

00:20:13: What's your view on our country right now and I will ask him tomorrow what you are saying,

00:20:20: I think that the race narrative is actually just a bit of wrong way to think about it, because when everybody thinks there's a race they only have one winner.

00:20:35: The reality with AI is that there are going be multiple winners from diffusion right, not necessarily from invention.

00:20:43: There may only be one company that invents it or one country that invets it first but the beauty of this software in general is a software wants to be free.

00:20:51: Software is something that can be diffused very easily because its you know when USB stick anywhere or one IP address.

00:21:00: So it's less important who invents and is more important Who diffuses into the marketplace?

00:21:08: And I think this This is something that Germany, actually all of Europe can really take solace in That you don't have to be the inventor.

00:21:17: In fact a lot of American Progress came from taking ideas that come from Europe or UK right and German.

00:21:27: they just diffused it better for a period of time.

00:21:30: And I think right now, a lot of the success in China actually came from ideas that come into other parts of the world.

00:21:35: They just defuse their better and China well Now when this technology is either invented in China or invented in The US.

00:21:43: Europe should just think about how can i diffuse if?

00:21:46: How Can I do It without crushing the economy Without creating you know social unrest?

00:21:51: and I Think That's something at Germany and Actually most Of Europe does Better than most places in the world, is that they protect their people.

00:21:59: They have policies around them and social safety nets.

00:22:02: I think That should be helpful.

00:22:05: And yesterday i've read a wonderful news The new invidia coming from Germany.

00:22:13: It's on the quantum computer field.

00:22:17: We are doing quite well there.

00:22:20: Biggest threat is Americans will buy it.

00:22:23: But I think too that, Matt in the quantum computer field there's a lot of things are going on right now.

00:22:30: Right?

00:22:31: Yeah!

00:22:31: I mean...I think quantum is an important technology but i also feel like it was a lot maybe misunderstanding how much impact quantum can have because Quantum is not faster computers and better computers.

00:22:46: It's a specialized computer which works at certain segments Of industry or certain types of calculations and right now it's a relatively narrow set of activities Or types of problems that they can solve.

00:23:01: And so, It will not replace Traditional GPUs or CPUs?

00:23:07: So I think Germany is getting these kind of breakthroughs.

00:23:12: everywhere in the world there are smart people so its gonna happen at lot places but i went over index the economic impact of this.

00:23:21: I think it will help in certain areas, and particularly probably things related to cyber and encryption or maybe stochastic forecasting but those are not necessarily going to lead to major economic impacts near term.

00:23:37: So leave us this news that we have two good news.

00:23:40: The first one is the new NVIDIA from Germany and second, we win the championship of soccer in the United States.

00:23:47: It's not sure but will try.

00:23:50: Well yeah!

00:23:51: The World Cup was coming so I hope you can think about it.

00:23:54: One other good news i think Europe Can Win the Diffusion Race That's the race should be focused on Not on the Invention Race But on the diffusion race.

00:24:03: One other point We didn't talk about.

00:24:06: This week, Trump is in Beijing talking to Xi Jinping and I think AI technology.

00:24:13: will be one of the big topics.

00:24:15: I think they said essentially it's that the three T is this tariffs, technology and Taiwan so... So uh i-I think a t-the-the technology side we'll be the ones that we care the most about.

00:24:27: And um you know I was fortunate enough to have some impact in terms of someone at the agenda for or the technology side.

00:24:36: So

00:24:38: what are the real good news in this?

00:24:40: Do you believe there will be a breakthrough and that with these two big guys, they'll end war in Irania or... What do you see?

00:24:53: I feel like it would definitely be a topic.

00:24:56: But i don't really have any view on political side.

00:25:02: but breakthrough in terms of the willingness to cooperate on technology, and maybe reduction some of the technology export controls.

00:25:16: And then there's more cooperation on safety.

00:25:19: so I think those are things that will help the world actually grow their economy.

00:25:24: And the

00:25:28: good news is they will meet each other in a few days, and it's fantastic because until a few months we didn't see that.

00:25:38: It couldn't be that they would cancel this right?

00:25:41: Yeah yeah I think next year maybe two or three more times.

00:25:45: so i'm optimistic.

00:25:49: there'll some positive traction.

00:25:53: We are already... The final question And the normal final question is what's on your mind right now?

00:26:01: and if there something that you want to share with us?

00:26:06: Yeah, I think recently i've spent a lot of time with folks in the defense industry.

00:26:13: There just very big misunderstanding that AI can or should be used as weapon.

00:26:24: We'll give a country a strategic decisive strategic advantage.

00:26:27: I think that's something that is creating Push for a lot of bad decisions or dangerous Decisions in the world and we really need to move away from that because though The more you understand them, the more you dig into how AI actually being applied In in the military.

00:26:44: it usually takes somewhere between one and two years Between the invention of a technology before it actually gets diffused into into Military And by that time, the commercial version—the industry version —that a consumer can use is actually more advanced than one being used in military.

00:27:01: We really should limit how much we position this as national security race for AI and focus on separating the two, where certain things let's not cooperate on.

00:27:17: but there are a lot of things.

00:27:18: Ninety-nine or ninety eight percent that can be done for the world.

00:27:22: AI can provide positive value like we said earlier all with public good.

00:27:27: whether you're talking about solving medicine issues making better materials and make companies more efficient these are all things which we should and can cooperate in and just kind of keep the military stuff separate from that.

00:27:40: And if we did that, I think world would be a safer place with more flourishing places.

00:27:46: That's dual use technology.

00:27:49: You can do it in bad way or very good ways.

00:27:53: We hope people will use it for better lives And let's talk about it.

00:28:00: It is incredible how much you know and what are your messages?

00:28:04: Thank-you very much for that, Arvind.

00:28:06: thank-you!

00:28:07: This was the Big Bang Tech Report but we will go on... We'll see again in a few days when we're happy if you switch on or accompany us.

00:28:20: What do you think is your last message to our viewers?

00:28:26: The world has huge positive potential.

00:28:29: You have to understand it and then take some action to support a positive path, so part of that is actually subscribing and watching us more often.

00:28:38: Very good!

00:28:38: Thank you very much and see you soon everybody!

00:28:42: Bye bye.